PB5: Contributed Session

Room: Old Main Academic Center 3110

Webex Link

Chair: Joonsik Hwang

Yassine El Marnissi, Mississippi State University

Time: 2:40 pm - 3:00 pm (CST)

Title: Microscopic structure analysis of diesel spray based on high-speed super-resolution image


The spray atomization and mixing process are key parameters for direct-injection engines. Understanding these processes is of utmost importance to improve engine performance and emission characteristics. This study investigated the effects of injection pressure on microscopic spray characteristics using a high-speed extinction image. The high-speed diesel sprays were assessed by liquid penetration, width, length of the intact core, and droplet diameter based on a super-resolution image enabled by a PDE-based interpolation algorithm. This method provided high resolution up to 1.5μm per pixel so droplet sizing could be performed. The experimental results showed enhanced spray breakup and air-fuel mixing with higher injection pressure.

(Joint work with Philku Lee, Seongjai Kim, and Joonsik Hwang)

Alexander S. Way, Mississippi State University

Time:  3:00 pm - 3:20 pm (CST)

Title: Vector Autoregressive Models for Predicting Time Dependent Solutions of Partial Differential Equations


Vector Autoregression (VAR) models are statistical models used to capture the bidirectional relationship between multiple variables that change over time. VAR models have become increasingly popular forecasting tools used in the fields of econometric and financial analysis and demonstrate applicability to data-driven modeling of physical dynamic processes as well. In this study, we investigate the use of VAR models to predict time-dependent solutions to partial differential equations (PDEs). The models are trained using only a subset of initial time series solutions to the PDEs, then used to forecast the solutions at future time steps beyond those of the training dataset. This technique represents a purely data-driven modeling approach and stands in contrast to a fully or partially “physics-informed” method, in which the models are encoded with a prior understanding of the physical process or governing equations. We demonstrate the application of VAR models to forecast the solutions to a variety of partial differential equations, including the advection equation, Burger’s Equation, and the wave equation.

(Joint work with Adrian Sescu and Ian Dettwiller)

Vance Hudson, Mississippi State University 

Time:  3:40 pm - 4:00 pm (CST)

Title: Full-Angle Computed Tomography (CT) of Gasoline Spray


The performance of gasoline direct-injection (GDI) engines is heavily dependent on injection spray and the subsequent air-fuel mixing process compared to the conventional port fuel injection (PFI) engines. As more complex conditions are considered in simulations, more detailed experimental data is required to validate the resulting spray predictions. In this study, Full-Angle Computed Tomography (CT) is used to create a reconstruction of the spray pattern from an automotive-grade gasoline injector for future validation of machine-learning and CT algorithms. The CT was carried out based on the high-speed extinction images taken with baklit configuration. The raw images were converted into projected liquid volume map that can provide quantitative information for CFD validation

(Joint work with Yassine El Marnissi, Kyungwon Lee, Sofiane Youssaoui, Michael Gibson, and Joonsik Hwang)

Vasile Marinca,  Politehnica University of Timisoara

Time:  4:00 pm - 4:20 pm (CST)

Title: Analytical solutions for unsteady flow of gas through a porous medium by using Optimal Auxiliary Functions Method


The aim of the present work is to propose an accurate technique to investigate the unsteady flow of gas through a porous medium. One example leads to the conclusion that our procedure is simple, easy to use and very accurate. The approximate analytical solutions are obtained for a nonlinear differential equation with variable coefficients. Our procedure is independent of the presence of small or large parameters and is based on an original construction of the solutions. The results obtained for nonlinear boundary value problem on semi-infinite domain are compared with the collocation method, finite-difference method and Wu’s method.

(Joint work with Bogdan Marinca and Nicolae Herisanu)

Bogdan Marinca, Politehnica University of Timisoara

Time:  4:20 pm - 4:40 pm (CST)

Title: Solution of Troesch’s problem via Optimal Auxiliary Functions Method


As was pointed out by Troesch, the problem in this study of two-point boundary value problem is inherently unstable. A new procedure is employed to determine an approximate solution to nonlinear differential equation of Troesch’s problem. Our technique does not depend upon small or large parameters and provides us with a convenient way to optimally control the convergence of the approximate solutions. For the sake of brevity, it is very important to point out that some optimal convergence-control parameters appear, which lead to a high accuracy. Our results are compared with numerical or exact results or with results obtained through the Homotopy Perturbation Method of Adomian Decomposition Method. The proposed technique does not need restrictive hypotheses.

(Joint work with Vasile Marinca and Nicolae Herisanu)